Evaluating Agricultural BMP Effectiveness in Improving Freshwater Provisioning Under Changing Climate

Abstract

Freshwater provisioning (FWP) is a critical ecosystem service that is highly affected by climate change/variability as well as land use/land management. Agricultural best management practices (BMPs) are implemented to mitigate the adverse impacts of intensive agricultural production on flow and water quality, thus can potentially protect and improve FWP services. Many studies have assessed BMP effectiveness for improving hydrology/water quality, however the impact of climate changes on BMP effectiveness for protecting FWP is poorly understood. In this study, changes in FWP under 5 BMPs and 6 projected climate change/variability scenarios, were quantified. The Soil and Water Assessment Tool was used to quantify FWP services for baseline (1975–2004) and future climates (2021–2050). We then assessed the climate change impacts on BMP effectiveness for 13 watersheds in the Upper Mississippi River Basin. The results indicated that all 5 BMP scenarios behaved similarly under the historical and future climates, generally resulting in improved FWP services compared to the baseline agricultural management. The combined BMPs was the most effective way to enhance FWP. No-tillage and cover crops performed well in improving FWP in agriculturally-dominated watersheds, while filter strips and grassed waterways had high effectiveness in non-agriculturally dominated watersheds. Results for the climate scenarios indicate that 5 BMPs under future climate were still effective compared to baseline. The increased precipitation and rising temperatures generally improved BMP effectiveness in maintaining and improving FWP services, due to increased freshwater availability under the projected future climate.

Introduction

Freshwater provisioning (FWP) is one of the critical ecosystem services identified by the Millennium Ecosystem Assessment (MEA) that supports human health as well as other ecosystem services (MEA 2005). Ecosystems can control the phase, quality and availability of renewable freshwater resources (Vörösmarty et al. 2005). The supply of freshwater describes the ecosystem modification of water used for multiple purposes, such as drinking, recreation, fisheries and survival conditions of aquatic biota, etc., in the case that water conditions simultaneously meet the environmental flow requirements and water quality standards (Li et al. 2016). With the increasing global population, we are putting more strain on freshwater supplies in terms of both its physical availability and chemical characteristic through our disturbance of the landscape. Much of this influence is negative and attributed to climate change/variability as well as agricultural management (Vörösmarty et al. 2005). Climate change may alter the hydrological cycle and impact both quantity and quality of water discharged from watersheds (Terrado et al. 2014). According to global climate model projections, mean surface temperature is predicted to rise by 2.6 °C–4.8 °C by 2100 in North America (IPCC 2013). More extreme precipitation events with longer duration are expected and result in floods, which may damage the agricultural production and urban infrastructure (Dakhlalla and Parajuli 2016). These concerns of changing water quantity and quality resulting from projected climate changes will subsequently affect FWP services in watersheds.

In addition, land management may affect FWP through modifying components of hydrologic cycle, such as overusing water supplies via irrigation, and through the degradation of water quality from field losses of fertilizers and pesticides to streams (Willaarts et al. 2012). Agricultural best management practices (BMPs), often considered as agricultural conservation practices, are sets of effective measures that widely adopted by producers to reduce nonpoint source pollution in agricultural watersheds, through reducing pollutants at the sources, decelerating pollutant transport, or processing the impacted waterbodies. Watershed managers usually use models to estimate the effects of BMPs in improving water quality, since their effectiveness cannot be tested in all situations (Arabi et al. 2008). These BMPs can be structural and nonstructural and are commonly used in combinations to improve water quality, soil health, and crop production (Chaubey et al. 2010). No-tillage practices can reduce soil erosion because of the increased crop residue cover on the soil surface and the intact root systems below ground to stabilize soil particles until the next planting (Waidler et al. 2009). Cover crops play a similar role as no-tillage by providing soil stabilization and cover during fallow times (e.g. winter). Filter strips along field edges and grassed waterways within field water flow paths can lead to increased infiltration and sedimentation by slowing runoff and plant uptake of nutrients (Arabi et al. 2008; Waidler et al. 2009).

The Soil and Water Assessment Tool (SWAT) has been demonstrated to be an effective tool to simulate climate changes as well as BMP implementation strategies for complex landscapes (Arnold et al. 1998; Gassman et al. 2014). Many studies have applied the SWAT model to evaluate BMP effectiveness in improving hydrology/water quality at different spatial scales (Villarreal et al. 2004; Secchi et al. 2011; Kaini et al. 2012; Demissie et al. 2012). Some studies have evaluated BMP effectiveness at watershed scale in combination with projected climate scenarios (Chaubey et al. 2010; Bosch et al. 2014; Panagopoulos et al. 2014; Jayakody et al. 2014; Dakhlalla and Parajuli 2016). Some suggest that BMP effectiveness changes with climate variability and may offset the benefits obtained from BMP implementation. Bosch et al. (2014) tested the effectiveness of three structural BMPs under future climate scenarios in the Great Lakes, indicating the increased need for agricultural BMPs as their effectiveness reduced under future climate. Dakhlalla and Parajuli (2016) evaluated the performance of three BMPs in attenuating the peak streamflow, and found the effectiveness of BMPs decreases under increased rainfall or CO2 concentrations. Chaubey et al. (2010) evaluated the BMPs in combination with climate scenarios in a pasture-dominated watershed and concluded that BMP combinations could reduce the pollutant losses, and weather variability significantly impact BMP performance. However, the aforementioned studies all primarily conducted BMPs effectiveness in improving hydrologic/water quality attributes. These attributes combined can represent ecosystem functions that contribute to ecosystem services (Logsdon and Chaubey 2013). To the best of our knowledge, no studies have evaluated BMPs effectiveness in improving FWP by utilizing an ecosystem service-based approach.

Given that current BMP implementation may not be sufficient to resolve future water quality issues and improve freshwater provisioning, it is important to examine BMP effectiveness under future climate scenarios. Our previous study found the diminished annual FWP in the past 20 years (1995–2014) in the Upper Mississippi River Basin (UMRB) (Li et al. 2016). However, we have yet to evaluate the best ways (e.g. BMPs) to improve it considering given current climate conditions, as well as under the influence of climate change. Expanding on results of Li et al. (2016), the main objectives of this study are to: (1) evaluate the effectiveness of BMPs in improving FWP at the watershed scale, and (2) estimate the variability of BMP effectiveness due to climate changes. To achieve these objectives, baseline management and 5 BMP scenarios (No Tillage, Filter Strips, Cover Crops, Grassed Waterways and combination of above BMPs) were implemented in the SWAT model, under both present and future climates. Future climate was projected by three Coupled Model Intercomparison Project Phase 5 (CMIP5) regional climate models under two Representative Concentration Pathway scenarios (RCP4.5 and RCP8.5). The main novelty of this paper is quantitatively analyzing the freshwater provisioning from an ecosystem service perspective by using ecosystem service-based approach. In combination with strong physically-based hydrology model and climate scenario analysis methods, we deeply evaluate BMPs effectiveness in improving FWP services by considering the climate change effects. The study results would expand the methods used for ecosystem service researches, and provide an insight for watershed or landscape managers to develop effective BMPs strategies in protecting and enhancing ecosystem services under changing climate.

Methods

Study Area

The UMRB is the headwater basin of the Mississippi River with a drainage area 492,000 km2 (Fig. 1), encompassing large parts of Minnesota, Wisconsin, Iowa, Illinois and Missouri (Delong 2005). Over 60% of the basin is cropland or pasture and the rest of land uses primarily are forest, wetlands, lakes and urban areas (Delong 2005). The major croplands are distributed in southern Minnesota and northeastern and central Iowa. With the agriculture and urban development over the past 200 years, the UMRB has become a key contributor of nutrients and subsequent hypoxia in the Gulf of Mexico (Alexander et al. 2000). The conservation in the basin has a long history and nearly all cropped acres have some kind of conservation practices in the UMRB. The structural practices such as buffer strips and terrace, are in use on 45% of croplands, reported by the Conservation Effects Assessment Project (CEAP 2012) with study period 2003–2006. Reduced tillage is common with no-till and mulch till accounting for 28 and 63% respectively. The cover crops were used with less than 1% as reported by CEAP (2012). The UMRB is not only a large floodplain ecosystem but also a busy waterway transportation system (www.umrba.org/facts.htm). More than 30 million residents rely on basin’s freshwater for drinking, recreation, industrial supplies, power plant cooling, etc. (Srinivasan et al. 2010). Despite its ecological significance, the ecosystem functions and services were deeply influenced by river modifications for navigation and flood control, land conversions for agriculture production as well as urban development (Alexander et al. 2000). Thus, this study focused on the UMRB, and more specifically, on 13 8-digit Hydrological Unit Code (HUC) watersheds that were evaluated in Li et al. (2016). A summary of these watersheds with primary land use types and baseline climate information are provided in Table 1.

Fig. 1
figure1

Location of UMRB displaying HUC8 watersheds and model calibrated streamflow stations

Table 1 Studied watersheds with the primary land use types and climate information

SWAT Model Description

SWAT is a physically-based hydrological model developed by United States Department of Agriculture (Arnold et al. 1998). The model can predict effects of climate and land management practices on hydrology, sediment, and agricultural chemical yields in large complex watersheds. Spatially, SWAT divides a basin into subbasins and subsequently into smaller homogeneous units called hydrologic response units (HRUs) (Neitsch et al. 2009). Multiple physical processes including water and sediment movement, and nutrient cycling can be directly simulated at the HRU scale. The hydrology simulation is separated into the land and routing phases of hydrological cycle, which are based on the water balance equation of soil water content (Arnold et al. 1998). The sediment yields are simulated by the Modified Universal Soil Loss Equation (MUSLE, Neitsch et al. 2009). The transformation and movement of several forms of nitrogen and phosphorus are simulated through a function of nutrient cycles (Arabi et al. 2008). Agricultural management practices affect the nutrient cycling and can be simulated by defining the appropriate management parameters for each HRU. Based on its global applicability and unique ability to evaluate agricultural BMPs, SWAT was selected for use in this study.

Model Parameterization and Calibration

Topography is one important SWAT input factor, and was represented by 90-m resolution digital elevation model (DEM) for estimating the landscape parameters (e g. slope). Nine years of Cropland Data Layer (CDL) datasets (2006–2014) from the U.S. Department of Agriculture (USDA)-National Agricultural Statistics Service (NASS) were overlain to generate a land use layer and crop rotations. The USDA Natural Resources Conservation Services (USDA-NRCS 2013) 1:250000 State Soil Geographic (STATSGO) soil map was used to identify soil characteristics. Historic daily precipitation and maximum/minimum air temperature observations for 440 weather stations from 1985 to 2014 were retrieved from the National Climatic Data Center (NCDC).

The model calibration was conducted with the use of SWAT-CUP (Calibration and Uncertainty Procedures) software (Abbaspour et al. 2015). The calibration and validation periods were from 1995 to 2005 and from 2006 to 2014, respectively. Details about model set up and calibration can be found in Li et al. (2017). Generally, the model had satisfactory performance for hydrologic and water quality predictions, supported by the high coefficient of determination (R2) values and satisfactory Nash-Sutcliffe efficiency coefficient (NSE) values (>0.5 per the criteria recommended by Moriasi et al. (2007)). The R2 for monthly streamflow were ≥ 0.7 for almost all stations during calibration and validation periods, and the NSE ranged from 0.51–0.83 for calibration period and 0.56–0.84 for validation period. While for St. Croix and Joslin stations, the NSE were marginally acceptable for calibration period and were < 0.5 during validation period (0.19, 0.4 respectively). This potentially relates to the effects of natural lakes or wetlands around these subbasins, which can attenuate peak flows. The current model has a low ability of simulating wetland physical process, thus may affect the hydrology predictions in these areas.

The R2 ranged from 0.54–0.75 and NSE ranged from 0.5–0.75 for sediment calibration period for the majority of stations. Although the NSE showed unsatisfactory (below 0.5 for all stations) during validation period, the R2 revealed an acceptable correlation between simulated and observed loads which were > 0.5 in four out of six stations. For total phosphorus (TP) and total Nitrogen (TN) calibrations, the R2 were consistently >0.5 for all stations during both periods. The NSE for TP calibration were > 0.5 in four out of six stations, while it was <0.5 for Hastings and Valley city stations with 0.37 and 0.16 respectively. It is partially due to the poor statistics of runoff and sediment in these subbasins during validation period. The NSE for TN calibration displayed bit lower than TP, which ranged from 0.2–0.6 for calibration period and from 0.21–0.64 for validation period. It is not unexpected to get monthly statistics of TN, TP and sediment not as consistently strong as those of streamflow, given the large scale of model application and the uncertainty about some input data such as the distribution of different tillage practices, fertilizer application time and manure nutrient inputs, etc. The monthly nutrient simulations might be greatly improved if more precise such management information can be collected through collaborations with stakeholders in such a large basin.

BMP Scenarios Establishment

Five BMP scenarios were implemented in SWAT including (1) No Tillage (NT); (2) Cover Crops (CC); (3) Filter Strips (FS); (4) Grassed Waterways (GW); (5) Combination of above BMPs (CM). NT practice can retain the crop residue on the soil surface and avoid the root systems disturbance below ground, which facilitate the accumulates of soil organic matters and increase the soil water permeability, thus reduce soil erosion and nutrients losses from fields to waters. CC practice provide ground cover for soil typically exposed during winter months by planting winter crop (e.g. cereal rye), and the crop residue covered on the ground will increase the soil texture and provide similar water quality benefits as NT. NT and CC were tested because both are commonly used conservation practices for water quality improvement in the Midwest (Kalcic et al. 2015). FS are vegetated areas between surface water bodies and cropland, in order to filter sediment, nutrients and chemicals from the runoff before entering rivers and streams. The sediment and nutrient transportation through FS was predicted through using the newest filter-strip routine in SWAT based on empirical solutions developed from the Vegetative Filter Strip Model (VFSMOD). Three parameters (VFSRATIO, the ratio of field area to vegetative FS area; VFSCH, the fraction of the flow through the most heavily loaded 10% of the FS which is fully channelized; VFSCON, the fraction of the total runoff from the entire field entering the most concentrated 10% of the vegetative FS) are major model inputs for FS simulation (Neitsch et al. 2009). GW are natural or constructed channels covered by vegetation, which can slow the water flow and minimize the channel surface erosion, thus reducing sediment and pollutants in water bodies (Waidler et al. 2009). CM scenario is a combination of all four BMPs at the same time in each cropland, in order to examine the synthesized effectiveness of these BMPs in improving FWP. The implementation of BMP scenarios in SWAT and management parameters setting were according to the SWAT model theoretical documentation and the Conservation Practice Modeling Guide (Neitsch et al. 2009; Waidler et al. 2009).

The NT scenario was implemented on corn-soybean and continuous corn rotations with conventional, reduced, or mulch tillage. The SWAT default NT operation was used which has low mixing efficiency (0.05) and passes with a lower depth (25 mm) at planting. The curve number (CN) was reduced by 2 points to account for increased infiltration in NT systems. Manning’s roughness coefficient for overland flow (OV_N) was increased from the default value (0.14) to 0.2 to account for a rougher surface. The crop factor (USLE_C) reduced from default value (0.2) to 0.05 representing a decreased ratio of soil loss from cropland to the corresponding loss from clean-tilled or continuous fallow land (Panagopoulos et al. 2014; Kalcic et al. 2015).

The CC scenario was implemented on corn-soybean and continuous corn rotations. CC were modeled as cereal rye, since it is a recommended winter cover crop in corn belt and the parameters are available in crop database. Cereal rye was planted in October after harvest of corn and killed in April before planting corn/soybean in the spring.

The FS and GW scenarios were also implemented on HRUs with corn-soybean and continuous corn rotations to keep the result comparable. It is recognized that implementation of BMPs with such a large scale in the basin would be unrealistic. However, these BMPs were just used for evaluating the resultant changes of FWP services and providing scientifically based estimates of the BMP effectiveness in terms of improving FWP under different climate conditions. The FS and GW were simulated at the starting warm up period. FILTER_I and GWATI were set to 1 to flag simulation. The fraction of the total runoff from the entire field to the most concentrated 10% of the vegetative FS (VFS) (VFSCON) was set to 0.5 (Waidler et al. 2009). The VFSRATIO was set to 40, assuming that 2.5% of field area was VFS area. GWATW (Grassed WATerway Width) was 10 assuming that the average top width of grassed waterway is 10 m.

Climate Model Description and Data Source

Three regional climate models (RCM): CCCma-CanESM2 RCA4, ICHEC-EC-EARTH RCA4 and ICHEC-EC-EARTH HIRHAM5 with two RCP scenarios were retrieved from Coordinated Regional Downscaling Experiment (CORDEX) simulations. The CORDEX provide high-resolution “downscaled” climate data based on CMIP5 (Coupled Model Intercomparison Project Phase 5) simulations (Taylor et al. 2012). The grid resolution of CORDEX models in North America is 0.44°×0.44°. CCCma-CanESM2 RCA4 and ICHEC-EC-EARTH RCA4 are both RCA4 models from Swedish Meteorological and Hydrological Institute (SMHI), and forced by CCCma-CanESM2 and ICHEC-EC-EARTH global models respectively. ICHEC-EC-EARTH HIRHAM5 model is from Danish Meteorological Institute and driven by ICHEC-EC-EARTH global model (Russo et al. 2016). The RCPs for CMIP5 are greenhouse gas concentrations scenarios that were established based on the projections of future population growth, technological development and societal responses. The RCP8.5 represents a high radiative forcing scenario with stabilization after 2250. The RCP4.5 is intermediate scenario with medium-low radiative forcing that stabilize after 2150 (Taylor et al. 2012). Considering the data availability, the RCP4.5 and RCP8.5 scenarios were utilized in this study.

The metadata of three models under RCP4.5 and RCP8.5, with time periods of 1951–2100, were downloaded from Earth System Grid Federation Node at the German Climate Computing Centre. Bias correction is usually needed as RCM simulations of temperature and precipitation often show significant biases such as occurrence of too many wet days with low-intensity rain or incorrect estimation of extreme temperature, due to systematic model errors caused by imperfect conceptualization, discretization and spatial averaging within grid cells (Teutschbein and Seibert 2012). Thus a distribution mapping method was utilized in this study for bias corrections of precipitation and temperature simulations, which is based on correcting the distribution function of simulated values to fit with the observed distribution function (Christensen et al. 2008; Teutschbein and Seibert 2012). A transfer is required to shift the occurrence distributions of precipitation and temperature. The bias-corrected daily precipitation and temperature data from 1975 to 2050 were used for analysis considering two 30-year periods representing baseline (1975–2004) and future (2021–2050) climate conditions.

Scenario Implementation in SWAT

Baseline management and 5 BMPs scenarios were tested in SWAT [(1) Baseline with no BMPs (BS); (2) NT; (3) CC; (4) FS; (5) GW and (6) CM] in order to evaluate both individual and combined BMP effectiveness in improving and protecting FWP. In addition to implementation under baseline climate, each BMP scenario was tested under six future climate projections as a function of three RCMs combined with two RCP scenarios. Thus, a total of 72 scenarios were evaluated to assess the BMP effectiveness in improving FWP services under climate change.

Results

Future Climate Change

The climate-model derived average annual precipitation and temperatures of baseline and future climate periods for each watershed are shown in Fig. 2. The average annual precipitation ranged from 753.7–1070 mm for baseline period, and ranged from 805–1112 mm for RCP4.5 and 823.3–1109.9 mm for RCP8.5 (Fig. 2a). The average annual precipitation increased under future climate compared to baseline for each watershed, ranging from 2.9%–8.1% under RCP4.5. While greater increased precipitation was indicated under RCP8.5 for most watersheds ranging from 3.7%–13.1%.

Fig. 2
figure2

Average annual precipitation (a) and temperature (b) for baseline and future climates according to the average of three RCMs; a-m in x-axis denotes each watershed (see Fig. 1)

Similarly, the average annual temperature under both RCPs had a general rise for each watershed compared to baseline climate (Fig. 2b). The temperature under RCP4.5 rose by 1.42 °C–1.68 °C, while for RCP8.5, the temperature increases were greater, between 1.58 °C–1.86 °C.

The average monthly precipitation and temperatures of baseline and future climates under RCP4.5 and RCP8.5 are provided in Figs. 3 and 4, respectively. The precipitation has a seasonal variability for studied watersheds (Fig. 3). Increased precipitation was predicted in spring (Mar-May) and autumn (Sep-Nov) for most watersheds under both RCP scenarios. However, reduced precipitation was predicted in summer (June–July) for 6 watersheds (Twin Cities, Lower Minnesota, Maquoketa, Lower Wapsipinicon, Upper Iowa and Lower Iowa) under RCP4.5 and almost all watersheds under RCP8.5. These results indicate that the future climate will get drier in summer and wetter in spring.

Fig. 3
figure3

Average monthly precipitation for baseline and future climates according to the average of three RCMs under two RCP scenarios; am denotes each watershed

Fig. 4
figure4

Average monthly temperature for baseline and future climates according to the average of three RCMs under two RCP scenarios; am denotes each watershed

Increased temperatures were predicted for all future months of each watershed (Fig. 4). Greater increases occurred in January, February and May compared to other months for most watersheds. It suggests that the UMRB will likely experience rising air temperature in the future. Comparing the temperature under both RCPs, a greater increase under RCP8.5 was predicted for most months, indicating a warmer climate predicted by RCP8.5 in the future.

BMP Effectiveness in Improving FWP under Baseline Climate

The annual FWP was calculated using an index-based method proposed by Logsdon and Chaubey (2013) that incorporates a metric of water quantity and a metric of water quality. Details of this method can be acquired from Li et al. (2017). The mean annual FWP under 5 BMP scenarios were improved, as expected, compared to baseline management for all studied watersheds (Fig. 5). The median and maximum of annual FWP under CM scenario also increased, although the corresponding minimum values had slight changes for most watersheds. The Q75 (upper quartile, meaning 25% of the data lie above this threshold) and Q25 (lower quartile, meaning 75% of the data lie above this threshold) values were little improved comparing to baseline management. These results further confirmed that the annual FWP could be improved either by implementing single BMPs or a combination of all.

Fig. 5
figure5

Boxplots of annual FWP under baseline management and 5 BMP scenarios for baseline climate; Explanations of BMP abbreviations in x-axis shall be referred to section 2.4; am denotes each watershed; Boxes denote the 1st and 3rd quartiles for annual FWP, the whiskers denote the maximum and minimum, and the red points denote the mean annual FWP

In addition, the mean annual FWP under CM scenario had an obvious greater increase than the other BMPs for all watersheds (Fig. 5). Great increases of median, maximum, Q75 and Q25 values were also observed under CM for most watersheds. It suggests that the annual FWP was further improved by combining BMPs than implementing them individually. The effectiveness of CM was highest among BMP scenarios.

However, the simulated effectiveness of NT, CC, FS and GW in improving FWP varied among watersheds (Fig. 5). The most effective BMP of each watershed is shown in Fig. 6. In the Lower Minnesota (b), Upper Iowa (h) and Lower Iowa (i) watersheds, the mean annual FWP and median, Q25 as well as Q75 values under NT scenario were greater than those under other scenarios (Fig. 5). It indicates that the effectiveness of NT scenario is relatively high in these 3 watersheds, suggesting this BMP might be of greater interest in these watersheds. The mean annual FWP under NT for these 3 watersheds increased by 11.4%, 10 and 10%, respectively.

Fig. 6.
figure6

A map of most effective BMP for studied watersheds (am); The color for each watershed denotes the most effective BMP; Explanations of BMP abbreviations can be referred to section 2.4

The effectiveness of CC scenario is greater than NT, FS and GW scenarios for 5 midstream watersheds (Turkey (c), Maquoketa (d), Copperas-Duck (e), Lower Wapsipinicon (f) and Skunk (g)). As shown in Fig. 5, the mean annual FWP, median and maximum values were greater under CC compared to other 3 BMPs. The mean annual FWP under CC for these watersheds increased by 16.5%, 24.1%, 12.6%, 9.8%, 9.9%, respectively.

The mean annual FWP was relatively high under FS for Twin Cities (a) and Lower Des Moines (j) watersheds, indicating high effectiveness of FS scenario in both watersheds. The mean annual FWP increased by 7.4 and 25% respectively. In terms of three downstream watersheds (Peruque-Piasa (k), Lower Illinois (l) and Upper Mississippi-Cape Girardeau (m)), the mean annual FWP, median, maximum and Q75 values were greater under GW than other BMPs, which suggests that GW performed best in improving FWP in these watersheds.

BMP Effectiveness in Improving FWP Under Future Climate

Changes in FWP of both the baseline management and BMP scenarios in response to climate change/variability, was quantified using the previously described projections of future climate (6 scenarios: three RCMs each executed with two emission scenarios). The average annual FWP under baseline management for baseline and future climates are displayed in Fig. 7.

Fig. 7
figure7

Mean annual FWP under baseline management for baseline and future climates of two RCPs; a-m in x-axis denotes each watershed; The y-axis displays as log scale

Comparing the annual FWP between baseline climate and two emission scenarios (RCP4.5, RCP8.5), it was improved under both RCPs for most watersheds (Fig. 7). The mean annual FWP under RCP4.5 increased ranging from 0.5%–38.5%, while decreases in FWP were indicated for Turkey and Peruque-Piasa watersheds (−0.2% and − 3.1%, respectively). The mean annual FWP under RCP8.5 increased for all watersheds ranging from 0.9%–40.9%. Furthermore, it increased more under RCP8.5 than that under RCP4.5 for almost all except Skunk watershed, which had an opposite trend. Overall, these results suggest that the annual FWP under baseline management will be improved in a future climate with rising temperatures and increased precipitation for most watersheds.

The annual FWP variations under baseline management and 5 BMP scenarios in a future climate are provided in Fig. 8. Comparing results under two climate periods, changes of annual FWP under 5 BMP scenarios in response to future climate followed similar trends to those under baseline climate (Figs. 5 and 8). The CM scenario showed consistently high effectiveness in improving FWP among all BMPs under future climate. The mean annual FWP and median, maximum, Q75 and Q25 values under CM all greater increased based on baseline management than those under the other BMPs.

Fig. 8
figure8

Boxplots of annual FWP under baseline management and 5 BMP scenarios for future climate according to average of three RCMs and two RCP scenarios; Explanations of BMP abbreviations in x-axis shall be referred to section 2.4; am denotes each watershed; Boxes denote the 1st and 3rd quartiles for annual FWP series, the whiskers denote the maximum and minimum, and the red points denote the mean annual FWP

The simulated effectiveness of NT, CC, FS and GW varied among watersheds under future climate, but showed similar trends to those results under baseline climate (Figs. 5 and 8). Specifically, the NT scenario was more effective in improving FWP for 3 agriculturally dominated watersheds (Lower Minnesota, Upper Iowa and Lower Iowa), whereas CC scenario performed best in improving FWP for another 5 agricultural watersheds (Turkey, Maquoketa, Copperas-Duck, Lower Wapsipinicon and Skunk). It can be evidenced by the more increased mean annual FWP under NT and CC scenarios for these watersheds (Fig. 8). In terms of FS and GW scenarios, both had low effectiveness in improving FWP for these agricultural watersheds, while they showed relative high effectiveness for some non-agriculturally dominated watersheds including Twin Cities, Peruque-Piasa and Upper Mississippi-Cape Girardeau.

Influence of Climate Change on BMP Effectiveness in Improving FWP

Relative changes of mean annual FWP under 5 BMP scenarios compared to baseline management for baseline and future climates are shown in Fig. 9. The mean annual FWP under both RCPs greater increased for each BMP scenario than those under baseline climate for most watersheds, particularly in Twin Cities, Lower Minnesota, Lower Wapsipinicon, Skunk, Lower Des Moines and Upper Mississippi-Cape Girardeau watersheds. In Upper Mississippi-Cape Girardeau watershed, the improvement of annual FWP for RCP4.5 under NT, CC, FS, GW and CM scenarios increased from 6.5%, 7%, 11.5%, 12.5 and 22.3% (baseline climate) to 48.8%, 48.7%, 55.3%, 56.8 and 71.9%, respectively. The improvement of annual FWP for RCP8.5 under 5 BMPs increased to 51.2%, 50.5%, 57.4%, 58.9 and 73.2%, respectively, which saw higher increases than those for RCP4.5. These results indicate that the effectiveness of 5 BMPs under future climate was improved compared baseline climate for most watersheds. Future climate change with rising temperatures and increased precipitation may enhance BMP effectiveness in improving FWP.

Fig. 9
figure9

Relative changes of mean annual FWP under 5 BMP scenarios compared to baseline management for baseline and future climate periods; Explanations of BMP abbreviations in x-axis can be referred to section 2.4; am denotes each watershed

However, in Peruque-Piasa watershed, the effectiveness of 5 BMPs decreased under RCP4.5 in comparison with baseline climate. The mean annual FWP under NT, CC, FS, GW and CM for RCP4.5 increased less (2.2%, 3.6%, 5.7%, 6.5 and 14.8%, respectively) than those for baseline climate (5.2%, 7.4%, 9.2%, 10.1 and 18.8%, respectively) (Fig. 9). Similarly, in Lower Illinois watershed, the mean annual FWP under CC, GW and CM scenarios for RCP4.5 increased less (2.9%, 5.8%, 9.9%, respectively) compared to baseline climate (3.1%, 5.9 and 10.9%), indicating the decreased effectiveness of CC, GW and CM scenarios under future climate. It potentially because the water quality condition was reduced under future climate resulting from increased water quality concentrations, the FWP service thus would be reduced under baseline management and future climate. This may indirectly affect the BMP effectiveness in improving FWP services for both watersheds.

Discussion

BMP Effectiveness in Improving FWP

The results of this study indicate that the annual FWP under 5 BMP scenarios would increase, as expected, compared to baseline management for all studied watersheds. This is potentially because total nitrogen, total phosphorus and sediment concentrations (CTN, CTP, CTSS) were reduced for most watersheds by implementing different BMPs compared to baseline management (Fig. 10). As a result, the water quality conditions would be improved resulting in enhanced FWP services. Our previous study showed that water quality is a more important and sensitive factor of FWP, so this is unsurprising (Li et al. 2016, 2017). Comparing the effectiveness of CM to single BMPs, CM performed better in terms of improving FWP among all BMPs for both baseline and future climates. This is primarily because nutrient concentrations under CM scenario were reduced to a greater extent compared to other BMPs, which may greatly improve the water quality and enhance FWP (Fig. 10). It should be noted though that a combination of BMPs across large amounts of cropland in the UMRB would be costly. Although one BMP is less effective, it may be more efficient with respect to cost of implementation.

Fig. 10
figure10

Relative changes of total nitrogen (TN), total phosphorus (TP) and total suspended solids (TSS) concentrations under 5 BMPs compared to baseline management for baseline climate; a-m in x-axis denotes each watershed; Explanations of BMP abbreviations can be referred to section 2.4

As shown in Fig. 10, CTP and CTSS under CM scenario had distinct reductions among watersheds ranged from 9.7%–36%, and 1.9%–67.6%, respectively. CTN under CM also reduced for almost all watersheds ranging from 0.9%–44.9%, except for lower Illinois (l) watershed. Similar results can be found in previous studies (Chaubey et al. 2010; Bosch et al. 2013). For instance, Bosch et al. (2013) concluded that greatest reduction of sediment and nutrient yields occurred when all BMPs were implemented simultaneously, through testing NT, CC and FS scenarios for six Lake Erie watersheds.

Comparing the effectiveness of the single BMPs, CC scenario performed the best in terms of improving FWP for 5 midstream watersheds (Turkey, Maquoketa, Copperas-Duck, Lower Wapsipinicon and Skunk), whereas NT scenario had high effectiveness in Lower Minnesota, Upper Iowa and Lower Iowa watersheds. All of these 8 watersheds are agriculturally-dominated watersheds with cropland area ≥ 50% (Table 1), which illustrates that CC and NT scenarios had high effectiveness in agricultural watersheds. This can be further explained by the great reductions of CTN, CTP and CTSS under both BMPs compared with baseline management. As shown in Fig. 10, CTN, CTP and CTSS under CC scenario in Maquoketa watershed reduced by 23.6%, 26.6 and 33.1%, respectively. CTP and CTSS under CC scenario also decreased in the other 4 watersheds, which may improve water quality conditions and enhance FWP services. As previous studies showed, the rye crop could effectively reduce nitrate concentrations in drainage water, and it is therefore a viable management option to reduce nitrate loads delivered to surface waters from agricultural drainage systems in the UMRB (Kaspar et al. 2012; Strock et al. 2004). This could be one potential reason that CC scenario behaved most effective in these agricultural watersheds.

Great reductions of CTN, CTP and CTSS occurred in Twin Cities and Lower Des Moines watersheds under FS scenario, and great reductions of those occurred in three downstream watersheds (Peruque-Piasa, Lower Illinois and Upper Mississippi-Cape Girardeau) under GW scenario (Fig. 10). As a result, the water quality would be greatly improved and enhance FWP services. Therefore, FS and GW performed well in improving FWP in these 5 watersheds.

However, the results of this study indicate lower effectiveness of FS and GW in agricultural watersheds (Lower Minnesota, Lower Des Moines, Lower Illinois, Turkey, Maquoketa, Copperas-Duck, Lower Wapsipinicon and Skunk), compared to NT and CC scenarios. It is likely due to small reductions of CTN in these watersheds under both scenarios (Fig. 10). As evidenced by previous studies, FS and GW may have little nitrogen reduction benefits in the tile-drained areas, potentially because excessive nitrate flows through subsurface tile drainage and bypasses the filtering through the FS and GW, resulting in decreased nitrate removal effectiveness (Bhattarai et al. 2009; Kalcic et al. 2015). As a result, relatively low effectiveness of FS and GW scenarios was indicated in these agricultural watersheds.

Influence of Climate Changes on BMP Effectiveness

Results from this study indicate that the effectiveness of 5 BMPs (NT, CC, FS, GW, CM) under future climate was improved compared to baseline climate for most watersheds. This is potentially related to the great increases in streamflow with smaller decreases in water quality under future climate, which may lead to increases in FWP services and freshwater availability. As a result, the BMPs were still effective in terms of improving FWP under future climate. The mean annual streamflow, CTN, CTP, CTSS under baseline and future climates of two RCPs were given in Fig. 12, and their relative changes from baseline to future climate were shown in Fig. 11. It shows that streamflow consistently increased by 4.7%–16.8% for RCP4.5, and 13.1%–27.3% for RCP8.5 for all watersheds. Although CTN, CTP and CTSS increased under both RCP scenarios, the streamflow increased greater at most watersheds (Fig. 11), resulting increased FWP under future climate for almost all watersheds (Fig. 7). For Lower Minnesota, Lower Des Moines and Upper Mississippi-Cape Girardeau watersheds, the decreased CTP and CTSS coupled with increased streamflow under both RCPs would improve the annual FWP. The increased freshwater availability under future climate would potentially be beneficial to enhance the BMPs effectiveness in improving FWP.

Fig. 11
figure11

Relative changes of mean streamflow, CTN, CTP and CTSS from baseline to future climate under baseline management for two RCPs; am in x-axis denotes each watershed

However, our results indicated decreased effectiveness of 5 BMPs under RCP4.5 for Peruque-Piasa watershed, and decreased effectiveness of CC, GW and CM for Lower Illinois watershed as well. This would be likely due to smaller increases in streamflow (7.2%, 4.7% for two watersheds respectively) and greater increases in CTP (7.6%, 6.8% respectively) under RCP4.5, which resulted in reduced FWP for both watersheds (−3.1%, −0.2% respectively) (Figs 7 and 11a). The reduced water quality conditions, coupled with little changes in streamflow are the primary reason of decreased BMPs effectiveness under future climate in these watersheds.

These results highlight the importance of balancing water quantity and water quality components of FWP services. Under present climate, the addition of BMPs into agricultural landscapes has the potential to improve FWP through the improvement of the water quality component of freshwater provision. However, under a future climate, we found large increases in water quantity with lesser increases in pollutant concentrations. In the freshwater provisioning index that we used, quantity and quality were given equal weight, therefore under a future climate we found increased FWP despite decreased water quality. The equations developed by Logsdon and Chaubey (2013) were intended to be flexible, therefore, if water quality is of greater concern than water quantity for stakeholders in the UMRB, this equation can be weighted to reflect this.

Conclusions

This study examined the effectiveness of 5 BMPs in improving FWP services in the agriculturally-dominant UMRB under baseline climate and 6 climate change projections (three RCMs each executed with two RCP scenarios). Results indicate that all BMPs (NT, CC, FS, GW and CM) behaved similarly under the baseline and future climate regimes, generally resulting in improved FWP services compared to baseline agricultural management. The CM scenario, the combination of all BMPs, would be the most effective way to improve FWP through reducing sediment and nutrients load compared to implementing single, though the economic feasibility of this was not considered in this study. NT and CC scenarios improved FWP services for highly agricultural watersheds, primarily due to reduced sediment and nutrient concentrations. FS and GW scenarios improved FWP for some non-agriculturally dominated watersheds (Twin Cities, Peruque-Piasa and Upper Mississippi-Cape Girardeau) whereas they were less effective in agricultural watersheds. Results also indicate that the 5 BMPs under future climate were still effective at improving FWP compared to baseline climate for most watersheds. This is strongly linked to the increased streamflow compared to correspondingly smaller increases in sediment and nutrient concentrations under the projected future climate. As a result, stakeholders should strongly consider the balance of water quantity and water quality when managing watersheds for freshwater provisioning services.

References

  1. Abbaspour KC, Rouholahnejad E, Vaghefi S et al (2015) A continental-scale hydrology and water quality model for Europe: calibration and uncertainty of a high-resolution large-scale SWAT model. J Hydrol 524:733–752. https://doi.org/10.1016/j.jhydrol.2015.03.027

    Article  Google Scholar 

  2. Alexander RB, Smith RA, Schwarz GE (2000) Effect of stream channel size on the delivery of nitrogen to the Gulf of Mexico. Nature 403:758–761

    Article  Google Scholar 

  3. Arabi M, Frankenberger JR, Engel BA, Arnold JG (2008) Representation of agricultural conservation practices with SWAT. Hydrol Process 22:3042–3055. https://doi.org/10.1002/hyp.6890

    Article  Google Scholar 

  4. Arnold JG, Srinivasan R, Muttiah RS, Williams JR (1998) Large area hydrologic modeling and assessment part I: model development. J Am Water Resour Assoc 34:73–89

    Article  Google Scholar 

  5. Bhattarai R, Kalita PK, Patel MK (2009) Nutrient transport through a vegetative filter strip with subsurface drainage. J Environ Manag 90:1868–1876. https://doi.org/10.1016/j.jenvman.2008.12.010

    Article  Google Scholar 

  6. Bosch NS, Allan JD, Selegean JP, Scavia D (2013) Scenario-testing of agricultural best management practices in Lake Erie watersheds. J Great Lakes Res 39:429–436. https://doi.org/10.1016/j.jglr.2013.06.004

    Article  Google Scholar 

  7. Bosch NS, Evans MA, Scavia D, Allan JD (2014) Interacting effects of climate change and agricultural BMPs on nutrient runoff entering Lake Erie. J Great Lakes Res 40:581–589. https://doi.org/10.1016/j.jglr.2014.04.011

    Article  Google Scholar 

  8. Chaubey I, Chiang L, Gitau MW, Mohamed S (2010) Effectiveness of best management practices in improving water quality in a pasture-dominated watershed. J Soil Water Conserv 65:424–437. https://doi.org/10.2489/jswc.65.6.424

    Article  Google Scholar 

  9. Christensen JH, Boberg F, Christensen OB, Lucas-Picher P (2008) On the need for bias correction of regional climate change projections of temperature and precipitation. Geophys Res Lett 35:L20709. https://doi.org/10.1029/2008GL035694

    Article  Google Scholar 

  10. Conservation Effects Assessment Project (CEAP) (2012) Assessment of the Effects of conservation practices on cultivated cropland in the Upper Mississippi River Basin. United States Department of Agriculture (USDA), Natural Resources Conservation Service (NRCS), USA

  11. Dakhlalla AO, Parajuli PB (2016) Evaluation of the best management practices at the watershed scale to attenuate peak streamflow under climate change scenarios. Water Resour Manag 30:963–982. https://doi.org/10.1007/s11269-015-1202-9

    Article  Google Scholar 

  12. Delong MD (2005) 8 - upper Mississippi River basin A2 - Benke, Arthur C. In: Cushing CE (ed) Rivers of North America. Academic Press, Burlington, pp 326–373

    Google Scholar 

  13. Demissie Y, Yan E, Wu M (2012) Assessing regional hydrology and water quality implications of large-scale biofuel feedstock production in the upper Mississippi River basin. Environ Sci Technol 46:9174–9182. https://doi.org/10.1021/es300769k

    Article  Google Scholar 

  14. Gassman PW, Sadeghi AM, Srinivasan R (2014) Applications of the SWAT model special section: overview and insights. J Environ Qual 43:1. https://doi.org/10.2134/jeq2013.11.0466

    Article  Google Scholar 

  15. IPCC (2013) Climate Change 2013: The Physical Science Basis. Working Group I Contribution to the fifth assessment report of the International Panel on Climate Change. Cambridge University Press, Cambridge, and New York. https://www.ipcc.ch/report/ar5/wg1/. Accessed 3 Nov 2017

  16. Jayakody P, Parajuli PB, Cathcart TP (2014) Impacts of climate variability on water quality with best management practices in sub-tropical climate of USA. Hydrol Process 28:5776–5790. https://doi.org/10.1002/hyp.10088

    Article  Google Scholar 

  17. Kaini P, Artita K, Nicklow JW (2012) Optimizing structural best management practices using SWAT and genetic algorithm to improve water quality goals. Water Resour Manag 26:1827–1845. https://doi.org/10.1007/s11269-012-9989-0

    Article  Google Scholar 

  18. Kalcic MM, Frankenberger J, Chaubey I (2015) Spatial optimization of six conservation practices using Swat in tile-drained agricultural watersheds. J Am Water Resour Assoc 51:956–972. https://doi.org/10.1111/1752-1688.12338

    Article  Google Scholar 

  19. Kaspar TC, Jaynes DB, Parkin TB et al (2012) Effectiveness of oat and rye cover crops in reducing nitrate losses in drainage water. Agric Water Manag 110:25–33. https://doi.org/10.1016/j.agwat.2012.03.010

    Article  Google Scholar 

  20. Li P, Chaubey I, Muenich RL, Wei X (2016) Evaluation of freshwater provisioning for different ecosystem services in the Upper Mississippi River Basin: current status and drivers. Water 8:288. https://doi.org/10.3390/w8070288

    Article  Google Scholar 

  21. Li P, Omani N, Chaubey I, Wei X (2017) Evaluation of drought implications on ecosystem services: freshwater provisioning and food provisioning in the upper Mississippi River basin. Int J Environ Res Public Health 14:496. https://doi.org/10.3390/ijerph14050496

    Article  Google Scholar 

  22. Logsdon RA, Chaubey I (2013) A quantitative approach to evaluating ecosystem services. Ecol Model 257:57–65. https://doi.org/10.1016/j.ecolmodel.2013.02.009

    Article  Google Scholar 

  23. Millennium Ecosystem Assessment (MEA) (2005) Ecosystem services and human well-being: synthesis. Island Press, Washington

    Google Scholar 

  24. Moriasi DN, Arnold JG, Van Liew MW et al (2007) Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans ASABE 50:885–900

    Article  Google Scholar 

  25. Neitsch SL, Arnold JG, Kiniry JR, Williams JR (2009) Soil and water assessment tool theoretical documentation version 2009. Texas Water Resources Institute Technical Report No. 406. College Station, TX, USA. 2011:1–618

  26. Panagopoulos Y, Gassman PW, Arritt RW et al (2014) Surface water quality and cropping systems sustainability under a changing climate in the upper Mississippi River basin. J Soil Water Conserv 69:483–494. https://doi.org/10.2489/jswc.69.6.483

    Article  Google Scholar 

  27. Russo S, Marchese AF, Sillmann J, Immé G (2016) When will unusual heat waves become normal in a warming Africa? Environ Res Lett 11:054016. https://doi.org/10.1088/1748-9326/11/5/054016

    Article  Google Scholar 

  28. Secchi S, Gassman PW, Jha M et al (2011) Potential water quality changes due to corn expansion in the upper Mississippi River basin. Ecol Appl 21:1068–1084. https://doi.org/10.1890/09-0619.1

    Article  Google Scholar 

  29. Srinivasan R, Zhang X, Arnold J (2010) SWAT ungauged: hydrological budget and crop yield predictions in the upper Mississippi River basin. Trans ASABE 53:1533–1546

    Article  Google Scholar 

  30. Strock JS, Porter PM, Russelle MP (2004) Cover cropping to reduce nitrate loss through subsurface drainage in the northern U.S. Corn Belt. J Environ Qual 33:1010–1016

    Article  Google Scholar 

  31. Taylor KE, Stouffer RJ, Meehl GA (2012) An overview of CMIP5 and the experiment design. Bull Am Meteorol Soc 93:485–498. https://doi.org/10.1175/BAMS-D-11-00094.1

    Article  Google Scholar 

  32. Terrado M, Acuña V, Ennaanay D et al (2014) Impact of climate extremes on hydrological ecosystem services in a heavily humanized Mediterranean basin. Ecol Indic 37:199–209. https://doi.org/10.1016/j.ecolind.2013.01.016

    Article  Google Scholar 

  33. Teutschbein C, Seibert J (2012) Bias correction of regional climate model simulations for hydrological climate-change impact studies: review and evaluation of different methods. J Hydrol 456–457:12–29. https://doi.org/10.1016/j.jhydrol.2012.05.052

    Article  Google Scholar 

  34. USDA-NRCS (2013) Soil Web Portal http://soildatamart.nrcs.usda.gov/. Accessed 27 July 2015

  35. Villarreal EL, Semadeni-Davies A, Bengtsson L (2004) Inner city stormwater control using a combination of best management practices. Ecol Eng 22:279–298. https://doi.org/10.1016/j.ecoleng.2004.06.007

    Article  Google Scholar 

  36. Vörösmarty CJ, Bos R, Balvanera P (2005) Fresh water. Ecosyst Hum Well- Curr State Trends Find Cond Trends Work Group 1:165

  37. Waidler D, White M, Steglich E, Wang S, Williams J, Jones CA, Srinivasan R (2009) Conservation practice modeling guide for SWAT and APEX. Texas Water Resources Institute Technical Report No. 399, Texas A&M University System College Station, Texas p 77843–2118

  38. Willaarts BA, Volk M, Aguilera PA (2012) Assessing the ecosystem services supplied by freshwater flows in Mediterranean agroecosystems. Agric Water Manag 105:21–31. https://doi.org/10.1016/j.agwat.2011.12.019

    Article  Google Scholar 

Download references

Acknowledgements

This study was funded by China Postdoctoral Science Foundation (Award No. 043206018), the U.S. Department of Energy (Award No. DE-EE0004396) and the USDA-NIFA (Award No. S1063). Great appreciation to Hendrik Rathjens for his technical help on bias correction of climate projection data. The authors also would like to thank anonymous reviewers for their comments and suggestions to improve the manuscript.

Author information

Affiliations

Authors

Corresponding author

Correspondence to Xiaomei Wei.

Ethics declarations

Conflict of Interest

The authors declare that they have no conflict of interest.

Appendix

Appendix

Fig. 12
figure12

Mean annual streamflow, CTN, CTP, CTSS under baseline management for baseline and future climates of two RCPs; a-m in x-axis denotes each watershed; The y-axis of top-left figure displays as log scale; Error bars denote the standard deviation of annual data series

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Li, P., Muenich, R.L., Chaubey, I. et al. Evaluating Agricultural BMP Effectiveness in Improving Freshwater Provisioning Under Changing Climate. Water Resour Manage 33, 453–473 (2019). https://doi.org/10.1007/s11269-018-2098-y

Download citation

Keywords

  • Ecosystem service
  • Streamflow
  • Water quality
  • SWAT